Implementation of smartwatch user interface using machine learning based motion recognition

Today's consumer wearable devices such as smart-watches and fitness trackers suffer from limited usability owing to a small touchscreen. In order to overcome the limitation, there has been a lot of research focused on expanding the interaction space outside the small touchscreen. In this paper, we propose a motion based user interface to generate six input events using the built-in 9-axis sensors on consumer wearable devices. First, we present and design an interaction concept followed by implementations of machine learning based motion recognition and applications using the proposed user interface.